If a federal court decides to shutter a data center in Memphis, Tennessee, it “directly threatens” the US government’s ability to protect our national security, according to the Trump administration’s chief digital and artificial intelligence officer.

The case began in April when the NAACP filed a class action lawsuit against xAI, Elon Musk’s AI company. The complaint alleges that xAI’s Colossus 2 data center violates the Clean Air Act. The data center includes a power plant in Mississippi with 27 gas turbines. Those turbines, which the nation’s oldest civil rights organization says are operating without the necessary federal permits, can produce significant pollution, which poses serious health risks for nearby Black communities in Tennessee and Mississippi.

The US government filed a motion on June 15, encouraging the court to dismiss the case. The Memphis data center powers xAI’s ability to provide AI to its customers, one of which the US government, Cameron Stanley, the Department of Defense’s chief digital and artificial intelligence officer, wrote in the filing. 

The US government uses a specifically created, government version of Grok, called Grok Gov Model. That AI tech is integrated into the Maven Smart System, an AI-powered military system created by mega-defense contractor Palantir and used by the US and NATO. The Maven system is used in “targeting, intelligence, readiness and recruitment,” Stanley wrote. 

While the defense department has struck deals with eight major AI companies, Grok is one of three AI enterprise providers that are “equipped to sustain mission-critical operations” for secret and top secret missions. Grok has features that other frontier AI models do not, Stanley wrote. For this reason, the government says, Grok AI is essential for national security. 

Stanley cited Grok’s prowess in Operation Epic Fury, the US-Israeli bombing campaign against Iran that began in February. US defense systems that have integrated Grok AI “enabled US forces to deploy over 2,000 munitions to 2,000 distinct targets within 96 hours during Operation Epic Fury, a testament to the greatly increased operational efficiency made possible by the Grok Gov Model,” Stanley wrote. 

Stanley doesn’t clarify how exactly Grok was used during Operation Epic Fury, and the Department of Defense declined to comment. xAI did not respond to requests for more information on how its Grok Gov Model works.

Community backlash against the rising flood of data centers has been swift and fierce, but this new development in the NAACP’s case raises a new question: If AI is used in warfare, are data centers essential for national security? 

AI is already being used in the military and defense industries. The Defense Department’s chief technology officer said in May that AI use in the agency was up 1,775% over the previous year; Defense Secretary Pete Hegseth said in a statement in January that the US will “become an ‘AI-first’ warfighting force across all domains.” 

While AI can be used for general business tasks — the kinds that you and I can use AI for, like data management and writing emails — defense leaders have said AI holds promise in improving surveillance, targeting and autonomous weapons. And the government’s spending reflects those goals, with the defense department adding a $13.4 billion line item to its 2026 budget solely for autonomous systems.

But there are still many unanswered questions about the role that new AI technology — and the data centers that power it — should play in defense and military operations.

The clash of environmental law and national security

In response to Stanley’s statement, the NAACP said it plans to continue to “stand up for democracy” against what it called bullying and authoritarianism. 

“At a time when the ultra-rich seem to be protected and supported by some of our government entities, it is important that polluting industries don’t get to benefit at the expense of the health of Black communities,” Abre’ Conner, NAACP director of environmental and climate justice, said in a statement.

There are a lot of concerns about the environmental impact that data centers have on the communities they’re moving into. Data centers require billions of gallons of clean water to cool servers, use a lot of electricity to keep them running and their outputs have been known to contaminate the surrounding neighborhoods. Nearby communities have been challenging data centers’ operational permits on these environmental grounds. 

Watch this: AI Data Center Infrastructure Plans Continue to Draw Controversy

01:58

The Environmental Protection Agency and the US government often participate in environmental litigation. But the government’s action here isn’t typical, Vincent Joralemon, law professor at the University of California, Berkeley, says.

“The more unusual thing here is the government intervening on the side of a private polluter to dismiss a citizen suit entirely,” Joralemon says. “As far as I know, this is the first time the federal government has intervened in a Clean Air Act citizen suit against a private company to argue the case must be thrown out.”

The Clean Air Act establishes a federal standard for the acceptable amount of emissions of hazardous air pollutants from stationary and mobile sources. Big projects like data centers need government permits to operate, which include confirmation that a company has implemented the “best available pollution controls.” A key point of the NAACP’s argument is that xAI doesn’t have these permits.

There is ample precedent for enforcing the Clean Air Act and operational permits that likely strengthen the NAACP’s arguments, Joralemon says. But the “novel” national security argument from the government could affect how the court resolves the case.

“Even a judge who finds a violation might hesitate to order a full shutdown [of the data center] if the government frames it as cutting power to defense-critical AI and could instead narrow the remedy or compliance obligations,” Joralemon says. Appeals, no matter the ruling, are likely, he notes.

National security concerns have won out over environmental ones in the past, like in a 2008 US Supreme Court case that found Navy sonar activity was permissible, despite the harm to marine life. These defense carve-outs make sense in principle, Joralemon says. 

But, like other experts, he’s “somewhat skeptical of all AI companies claiming national security when it is convenient.”

AI in the military

The military’s adoption of new technology — from land mines to nuclear weapons to drones — has long stirred fear and raised the stakes for warfare and its aftermath. The advent of AI has turbocharged the debate in recent years. 

Experts raised concerns at the beginning of the year when the Pentagon got into a public feud with Anthropic after the AI developer refused to allow its Claude AI to be used for surveillance and autonomous weaponry. 

President Trump slammed the company, calling it “woke,” and since then, the Defense Department says it has transitioned two-thirds of its AI activities away from Anthropic’s tools to other models. OpenAI, which makes ChatGPT, picked up a Pentagon contract shortly after Anthropic bowed out.

Governmental AI can be used “anywhere from the back office to the battlefield,” says Michael Horowitz, political science professor at the University of Pennsylvania and former deputy assistant secretary of defense.

“There are military uses of AI for logistics or HR that look like the uses of any company in America, all the way up to potential uses of AI closer to the battlefield in helping commanders more quickly and accurately identify potential targets,” Horowitz says.

AI has been powering some government activities for a while now. Defense employees using the Maven Smart System use almost 2 billion tokens every day, Stanley wrote in his statement. Tokens are a basic measuring unit for AI processing; having a chatbot write a one-paragraph email might take 400 tokens, for example. Vibe coding an app would take significantly more, because it’s a more involved and compute-intensive process.

While Grok is one of three AI models being used to do this work, it’s unique in the AI industry. Grok is known for having lax safeguards compared to those applied by other AI companies. It made millions of sexual deepfakes earlier this year, and before that, it spewed racism and anti-semitism, even referring to itself as “MechaHitler.” The government’s Grok may be in a different harness than the one used to abuse people online, but it’s based on the same commercial architecture.

Stanley didn’t explicitly say Grok or other AI was used in targeting and missile deployment during Operation Epic Fury, but an AI-powered version of that future isn’t out of the realm of possibility. These systems need to have stringent safety guardrails. We’ve already seen how fatal it can be when these operations go awry. 

In March, shortly after the Pentagon’s dust-up with Anthropic, the US launched a missile strike against an Iranian elementary school, killing 156 people, 120 of whom were children. In the days following the strike, there were concerns that AI was used in the mission. An early analysis from the New York Times found that it was outdated location information that led to the deadly mistake — human error, not AI. The US’ internal investigation is ongoing, but members of Congress have been demanding more info on the role of AI in the strike from Hegseth and the department.

Having humans in the loop to make decisions about military activity will be essential to any AI-powered defense system. 

“We never want to be in a condition where AI is making the decision about the use of force — that’s up to commanders and operators,” Horowitz says.

If AI is essential, so are data centers

If AI is part of modern war, the energy sources powering it are essential — that’s the argument the government is making in its defense of xAI. Data centers must be recognized as “a long-term strategic tool vital to maintaining our technological advantage against adversaries,” Stanley wrote.

Right now, these data centers are powering all of our AI activity. The demand for AI is so high that AI companies are racing to build more — there are over 900 operational data centers in the US now, with over 1,200 more planned, according to data center tracking firm Clearview. All this energy is needed for AI companies to build and train new, more advanced AI models for all of us to use. 

The Trump administration is investing in using AI for its military operations, and data centers are a key piece of that puzzle. Because of that, “It’s likely that [data centers] will be thought about in the same breadth as critical infrastructure,” Horowitz says, similar to the electrical grid, cell networks and oil refineries.

If these AI systems and data centers are essential to national security, then there is an argument that they deserve government oversight, Joralemon says. But so far, the Trump administration has been hesitant to enact rules or laws around AI, lest it slow down American innovation compared to Chinese AI labs. 

So if warfare is getting an AI update, powered by data centers and potentially at the expense of the environment and the health of communities, then the systems need to work. The success or failure of these systems will have grave consequences.

“No one wants their technologies to work more than militaries because technologies that aren’t safe don’t work, and…it’s soldiers whose lives are on the line,” says Horowitz. “So the American military should be highly incentivized to get the integration of AI into military operations right.”